Proton treatment-plan dose verification with a DPM-based Monte Carlo engine

Introduction Independent proton dose verification requires a calculation pathway that is accurate enough to identify clinically relevant discrepancies while remaining practical to commission and operate. We use a pre-release version of the commercial RadMonteCarlo, a multithreaded CPU-based proton Monte Carlo engine built on a dose planning method (DPM) framework for independent treatment-plan recalculation. In this study, we evaluate the sensitivity of RadMonteCarlo to the conversion of CT Hounsfield units (HU) to relative stopping power (RSP). Methods RadMonteCarlo was commissioned from integrated depth-dose (IDD) curves, spot profiles, and absolute-output data for a pencil beam scanning proton system. Three HU-to-RSP assignments were evaluated: a density-derived parameterization, a Bethe–Bloch-based calculation, and a Schneider stoichiometric tissue model. Four anonymized clinical plans (prostate, bladder, lung, and brain) were recalculated. RadMonteCarlo dose was compared with the RayStation Monte Carlo dose, using three-dimensional gamma criteria of 3%/3 mm and 2%/ 2 mm. RadMonteCarlo distributions were also compared directly to isolate RSP sensitivity from TPS agreement. Results The commissioned fits achieved a mean R² of 0.9998, and re-simulated R80 values agreed with the measurements within 1 mm (mean difference +0.39 mm). All 12 clinical plan–RSP comparisons met the specified 95% criterion at 3%/3 mm. In direct RSP model comparisons, target mean-dose differences ranged from−1.15% to +1.20%, distal R80 shifts along the beam axes ranged from−0.3 to +1.9 mm, and the 95th-percentile absolute voxelwise dose difference ranged from 1.8% to 4.1%. Discussion These controlled comparisons demonstrate that HU-to-RSP selection produces localized, anatomy- and beam-path- dependent changes in the individual cases examined. RadMonteCarlo therefore provides a stable independent recalculation framework in which calibration-driven effects can be identified and interpreted separately from transport or source-model differences.